 Hello, welcome everybody. This is Sartre from LS Ntap. We have a training today coming up on translation memory and machine translation. But before we jump in, I'm going to take a second, let a few more people log on and introduce you to our new website. LS Ntap has just moved over to the D-Law template. We have guides on call center technology and online triage that are being highlighted and all of our trainings are now under the events tab. We've got about 12 more trainings coming up this year, including two on information security, one on disaster planning, one more on language access. So I definitely recommend checking those out. Also our YouTube channel has started to increase the number of videos we're putting on there. We're putting on some shorter videos that are focused on trying to take a five or ten minute takeaway from some of our longer webinars. For example, we have a full webinar up on web accessibility that LSE put on about a month ago. We put together ten tips for web accessibility that's about a quick five minute video to pull out some of those things. So I definitely recommend checking out our YouTube channel. So we're going to talk about tools to create language target language in content and it's all focused on written work. We're not going to talk about oral interpretation and I'm very happy because we have an amazing panel and we're going to focus first on the most well-known one which is machine translation and then we're going to go to translation memory. So each of the speakers will introduce themselves. What I wanted to do is, I don't know if you have seen this very helpful map from the self-represented litigant network, they have this map that is based on American survey data that looks and you can, this is just a print screen, if we were in the actual tool, you could zoom in and see what your counties look like and you can see that the gradient goes the highest one, the darkest one is for 38 to 94 percent. You can see that vastly populated areas of the country are having increasing language diversity and then the lightest one is areas that only have less than three percent of people that speak a language other than English in their home and you can see that those tend to be in areas where there's not large populations. So this is just a tool that I think it's really helpful to review and plan when you're making all sorts of decisions about providing services and what materials to put out because you can dig on this a little more. So just, you know, the diversity is basically increasing of the languages that people speak at home and just as a reminder that the standard that everybody that's required to provide language assistance is required to meet is a meaningful access standard and that's a that's a pretty important concept because it means that the language has to be understood, the translated language has to be understood by the group reading it and so if it is incorrect or grammatically incorrect the meaning could change and it won't be meaningful so just that's kind of the the standard there on either oral written or focus here is on written and the other thing I wanted to share is that we're going to focus on written we generally when we talk about language access people talk about interpreters and translations as if they were the same skill and we often go back and forth and we generally say the translator did this or the translator said that when you hear translation and oral said spoken word that's incorrect the technical word translation denotes the written language the technical word interpretation denotes oral and they are not interchangeable skills somebody could be a fantastic oral interpreter but they may not have the capacity or ability to translate and same thing though it's it's harder to translate than to speak but there could be cases where somebody is a fantastic translator but will not be comfortable just jumping cold into any conversation particular one of legal nature so just wanted to flag that the focus today is mostly on the written word not the spoken word and on machine translation I think that this is one that is more well known and understood I think it has captured our imagination and it's something that a lot of people use and either on phone apps or through the google translate probably the most common one just wanted to call out that there's a lot more and there's all sorts of choices of tools with this and so we're going to focus on this one Diana Glick who's going to be our next presenter will cover more on this from her perspective and experience and then the second half of the presentation we're going to look at translation memory so I'm going to give presenter status to Diana hold on just a second and we'll get into into the nitty gritty just one minute okay wish to you can you see my screen mm-hmm okay so let me go over here and good morning or good afternoon everyone okay now can everyone see my screen yes oh wonderful okay so um thank you Claudia it's um really great to be here with everyone this morning talking about one of my favorite topics in the world which is the intersection of law and language which I like to say that's where I work is that that intersection at the judicial council I am an attorney here with the Center for Families Children in the courts with the judicial council of California I am a former professional translator which is what I did before I went to law school and I'm a former Spanish teacher I'm also a current Spanish teacher I teach legal Spanish at UC Davis School of Law in Davis California and with my legal work here at the judicial council I really focused on lots of things related to self-represented litigants but I also have a big focus on limited English proficient court users and language access issues and so that is why I like to say that I work at this intersection of language and law and it's truly a joy because those are both great passions of mine and there's they have a lot of relationship and so I guess I'm here to talk about the bad news I'm not sure why I'm first but hopefully it won't all be bad news but the title of my presentation is the perils of Google Translate so it's not exactly all good news either because of the role that I have here at the judicial council of having formally been a translator and having the Spanish knowledge I do review all of our Spanish assets for both linguistic and legal accuracy and because of that I often get requests like the one that you see in this email here and I call this the dreaded email this happens I don't want to say frequently but it does happen and you can see here that this person is asking me to review a translation that she has already put through Google Translate so what they needed for this particular situation was they were doing a pilot program and they needed a few signs obviously here you can see they were just going to do some paper signs it's just a pilot program so they weren't going to invest in a lot of signage at this point but they needed a few paper signs and I'll just say that you know this is kind of a frustrating situation to be in as a professional but I will say that having experienced this several times throughout my career I would say that it's given me a lot of perspective on the way that Google Translate works and doesn't work and also it's taught me a lot about what has to be done with source text so when you are going to translate something most of the work actually happens before you even hand it over to the translator so that's a little bit of what I'm going to talk about today so as I said this person just needed a quick sign I just need a quick sign they weren't going to invest a lot of money it's just going to be on an eight and a half by 11 sheet of paper and so the first sign that they wanted translated was this way to find through so you can see that Google didn't do so well with that translation even if you don't have a lot of experience with Spanish you can probably tell that the Spanish is not great on this one if that's the only way to find that's not really going to work so why didn't this work um well first of all we have a couple of problems here first of all um which way right when when you say this way what are we talking about exactly um and also finds room you you can't imagine how much time I spent researching this issue of what courts call this place a finds room and this place is really where someone goes to usually to pay a traffic ticket um and so it turns out that there are a handful of courts that call it the finds room most people call it other things like payment desk or cashier or um finds desk maybe but the finds room that was a new one for me so I spent a lot of time researching that um to make sure that we were giving them a proper translation um so you know what did I learn from this experience well when you want to put together a sign really the key here is to use some visuals and our friends at transcend who we're going to hear from in a minute have developed a great set of legal icons that they share freely with the world and it's just a wonderful contribution to our work um they are the ones who developed the the little um people that you see there with the dollar sign um and also getting rid of the language of this way and just putting an arrow simplifies things a lot um and then I also played around a little bit with the finds room concept and I always find that this is um it can be a little bit tricky right because if you have a bronze plaque that says finds room over your finds room um maybe you need to stick with that and not try to change it um even though something like payment of finds or you know finds just finds alone might be a better way to go so what are some of those best practices um when we're talking about signage specifically first of all check for existing resources right because there is just no need to reinvent the wheel on multi-lingual signage at this point a lot of places have developed signs in different languages and in particular here in california we have a glossary of signage terms and icons it is the second to the last link on this page and that is located in our language access toolkit and it contains 73 of the most common terms that you will see in courthouse settings um so wayfinding and um directional signage and things like that um and the basic components of a courthouse the jury room the the clerk's office things like that so we have a lot of those things already translated and they've been put into plain language and translated into 10 languages so so much of this work has already been done so anytime someone comes to me with a signage issue i like to point them in this direction first um there's just no need to reinvent the wheel on signage so here's another example i just need a quick sentence i'll hear um and this quick sentence says unlawful detainer cases will be heard in room four simple enough but you can see what happened here in spanish um that became los casos de detención ilegal se escucharan en la sala cuatro so one way that you can check if you don't speak the language that you're translating into and i only speak one of them so um i have to employ this often is to back translate things to see what you're getting right um and when we back translate this we see that google is translating this as cases of illegal detention so those of us who work with the term unlawful detainer day in and day out this probably just blows right by us um and i don't even know whether that's a common term in other states but in california that is an eviction notice um you know a complaint for unlawful detainer leads to an eviction um or can but it certainly has nothing to do with illegal detention so you can see here where things really went awry with google um and why is that the case well okay first of all we have a legal term of art in unlawful detainer um this seemingly simple sentence is actually loaded with legal terms of art and some confusing constructions um and first of all unlawful detainer kind of shows off you know some of the etymological background of our legal terms that we use um some of them come from old english some of them actually derived from latin in this case from french and ultimately from latin um so you know very word nerd stuff here going on um but not so easy to translate it is a term of art um even the term hearing is actually a little bit of a term of art not all languages connect court processes with the term hearing um which has other connotations in english as well um also the construction of will be heard is a little bit confusing right i don't know if this is a sign or if this is on a piece of paper but um it could be a whole lot clearer and so why is this so difficult well in english context is everything all right so here is my cute little mouse in this beautiful field and my instructions are use the mouse to navigate through the form fields and enter the information in each one um in english we tend to use the same words to convey different concepts um sometimes many different concepts can be packed into a single term and how do we figure out what we're talking about well it's all about context so um you know court is another example of this the word court can have many different meanings in english and if you were to translate that in isolation you really wouldn't know what you were talking about so looking at an example that's not spanish for once um simplified chinese and i did have this checked with a chinese speaker colleague of mine here um you can see that in chinese they're not so much like the english in that respect right they have different words for different things um so here you have mouse and the term for mouse in chinese and what's interesting is that it appears my understanding from this talking to my colleague here is that this term in chinese is computer mouse right which which makes sense so they added a little bit to it to give it that context and to be able to understand what we're talking about um but for fields you really are talking about two very different words you have the field of wheat and then you have a form field completely different terms in chinese so the benefit that you get from having a human translator as opposed to a google translator a google translate program is that that person can understand or ask you know what kind of document is this what is the larger message of this document how will the document be used who is the intended user of this document who's going to be the reader um when you send a document to a human translator they're able to consider and understand all these factors or ask questions and really get a sense of the overall project so therefore my best practice is on this one in terms of legal concept and how to convey all of those would be to absolutely uh we recommend a professional translator um whenever you can um and we also recommend an additional an additional legal review step to be performed by a bilingual attorney um so this is important as well because sometimes something can be beautifully translated but not make sense within um this is what i've seen at the judicial council that something can be beautifully translated but not necessarily um be coherent with the way that we translate concepts at the judicial council and the way that the judiciary has chosen to translate and convey certain concepts so that legal review step is really critical to making sure that you have a a readable well translated and legally accurate document so here is a request i have not received by email i don't want to misconstrue that um but i just need a quick consent to an otherwise illegal search um i didn't get an email about this one but i was asked to share a little bit about this case it's a recent federal case um it's usb omarcu samura i'm happy to send the citation anyone who wants to read it it's out of the u.s district court for the district of canvass um and i'll just kind of give you an overall view of the facts of the case so this occurred on september 21st of 2017 three in the morning officer ryan wilting of the canvass highway patrol stopped mr omar cru samura on i-70 mr cru samura was driving with a suspended registration um because he spoke spanish and didn't speak or understand english very well the officer began to use google translate on his in-car laptop to communicate with um mr cru samura so at some point in the conversation officer wilting typed into google translate the fallen question can i search the car here is what was shown to mr cru samura as the spanish translation of the question you can see it's right here puedo buscar el auto after some back and forth indicating a lack of understanding mr cru samura said sure okay yeah that's a direct quote from the case the car was searched and drugs were uncovered during the search apparently quite a bit of drugs so cru samura filed a motion to suppress the evidence uncovered during the search on the ground that his consent was not freely given with full understanding because of the language barrier and the use of google translate which generated an unclear translation without his consent the search was illegal this was his argument so you can see here how it was translated um let's back translate this one because i find it fascinating first of all um those of you who speak spanish will spot the problem right away with searching being translated as buscar okay um buscar means to look for something not to look through something so if i'm going to help you look for your contact that fell out of your eye you would use buscar but if you were going to be looking through something you would have different words which are suggested over here such as registrar in section at some other terms in spanish um but what kills me about this whole thing is that when you back translate it gets even worse because it says can i find the car which is i mean that's completely wrong so yeah i don't know what google was thinking on that one um so what does the court say about all this um they called in experts they called in translators and um their decision was that they found it was not reasonable for the officer to rely on google translate for the purpose of seeking consent to search this man's car they suppressed the evidence of the drugs found during the search um specifically they found that consent was not obtained the translation was not precise the good faith exception did not apply in this case because apparently the officer had an ability to call into a dispatcher for language assistance um because it's spanish i think they had somebody sort of at the ready for him and he opted instead of you know using that resource he decided to pull up google translate and use google translate instead so ultimately the situation here is that those drugs were suppressed and um he got his motion suppressed granted so just some concluding thoughts you know is it ever okay to use google translate um i certainly have used it um i as a soccer mom have to communicate with people who speak many different languages sometimes let them know where the pizza party is when picture day is um i use it for informal communications where we're not talking about life or death or drugs on the highway um for general understanding it can be very helpful um when you are in a complete bind i think that would be potentially a situation that might justify using google translate um i can't speak to any of the other tools out there i've only had this experience um really working with google translate and testing it for our own purposes um but i will say that you know the court in their decision in the cruz amora case did sort of leave it out um because of the fact that he had this access to um an interpreter and he could have called into his dispatcher and gotten access to someone um if he had not had that ability or perhaps if mr cruz amora spoke a language that he was not able to figure out or potentially um a language that's very rare and they didn't have an interpreter available to speak to him maybe that would have flown i don't know um but on the legal side i think that's kind of where it's down so the major takeaway is that i just wanted to share with everyone today is that um there are first of all there's a lot of resources out there if you need to translate something it just it's always paced to look to see what's out there first and most people are really happy to share their resources with you um including california so please do that and then if you find that you absolutely do need to create something completely from scratch there's a lot that you can do and really should do to facilitate translation including plain language review simplifying your structures using visuals all those things depending on what your translation is and at the end of the day it's really hard to go wrong with a professional translation um and then a best practice on top of that would be to have a legal review of the translation so that's it for me i don't know are we taking questions here or at the end caldia um we can definitely take questions here if we have them we all the questions that we have currently are attack oriented trying to help people out with stuff so we don't have any subject matter specific yet so caldia should i yeah yeah i'll um our next presenter is gonna be lilan samson and lilan i see that you have your mic ready to go and we are do you want to be the presenter or do you want me to run the slides for you um it's probably easier for me if you run the slides caldia okay i just have a couple here yeah so let me just do that i'll grab it then okay and just prompt me to change it i already sent a request to lilan to be a presenter all right let me see if that works i i selected um said okay i'm not sure if you can um i have them here if you want me to run them i um it says that lilan's presenter um did you go to full screen mode lilan now i think i got it yeah yeah okay that works okay so hello everyone my name is lilan samson i am an attorney at the maryland state law library and i run a web uh legal self-help website called the people's law library and claud you can go to the next uh so the people's law library is a legal self-help website the objective of which is to help people do legal research about issues that they might encounter in maryland state courts and so the objective is to provide plain language and legal information to help folks do legal research and solve problems and we can go to the next slide so what are some of the things that we provide we provide instructional articles topical articles and a couple of other services so in terms of instructional articles they're procedural type articles that describe how to do something so if someone needs to file a motion or file a case or file an appeal we have instructional articles on how to do the do that we have about 350 articles on substantive law that are sort of uh encyclopedia type articles some are as short as 500 words some are as long as a few thousand words we're in the process of taking those few those few articles that are in the thousands and and breaking them down into much more manageable chunks we also have a legal services directory that has over 100 different organizations in maryland that provide self-help resources for folks so if individuals are looking for pro bono or reduced cost legal assistance then they can find the appropriate organization in our directory we also have a clinic calendar that has a calendar of free legal clinics around the state we also help connect people with people so we have a the maryland court judiciary offers a self-help center where folks can chat with an attorney for free we have a link to the chat with a lawyer on all of our pages as well as ask a law librarian so folks can submit questions directly to the maryland state law library reference desk and have their questions answered directly from there and we can go to the next slide so what you're really here and interested in discussing is the way we use translation memory and how we translate articles so our workflow generally is this so someone a volunteer contacts me and i verify their credentials make sure that they have a proficiency in the foreign language as well as some legal credentials they either select an article or i assign them an article to translate that article is then uploaded to the translation management system that we use called lingo tech and the article or lingo tech interfaces directly with the people's law library content management system so that the article goes directly from our cms up to lingo tech lingo tech does its analysis of the article splits it up into segments and then allows the volunteer to actually get in and translate once the translation is completed then the article is assigned to a reviewer who is a licensed attorney who has the foreign language expertise as well following that review then the translation is re-imported into our content management system and it's published from there and we can go to the next slide so we have about the the languages that we support are spanish chinese korean and french we have over 150 articles currently translated in spanish we have our 100 most popular articles translated and that represents about 75 percent of all of our site traffic so the the breadth of coverage of the articles that we do have translated is is pretty broad and we can go to the next slide from here this is what lingo tech looks like from the volunteer perspective when they are doing the translation so the on the left hand side there is where the original text appears as a segment and the volunteer types in their information or types in their translation from there on the bottom it will fill in the spanish or the whatever the target translation is as the volunteer is going so they can have context and then on the next slide we'll demonstrate the translation memory so claude if you could uh the this is the translation memory so for individuals who are not familiar with it i'll try and give a brief description of it and i'm sure one of our later presenters can correct me if i get anything wrong here the translation memory takes previous translations that have been done and does an analysis of the thing that is trying to be translated now so it tries to do matching so if a word or a phrase or a sentence has previously been translated then it will provide that information to the translator so that the translator can do a translation that is consistent with what's come before so claude i think that finishes my presentation oh i'm sorry i have some benefits and some drawbacks of the translation memory system that we've used thus far so we find that it's it's helpful for encouraging volunteers people get really excited when they're able to use a really cool nifty uh online tool it's useful because it doesn't involve emailing documents back and forth and things getting messed up with formatting and losing links and things like that so it's it's very easy from on the volunteers to very quickly get started and reduces their management burden and their overhead the the system that we use also provides a custom glossary support so we were able to upload the glossary that we had previously used for translation for consistency purposes and the automatic segmentation is nice because it breaks up the article into manageable chunks so now we can discuss some of the drawbacks that we've encountered with the system the primary drawback that we've seen claude if you wouldn't mind going to the next slide yep there you go it's can be expensive in certain cases and the quality of the translation sort of varies based on the interpreters that you're using and in our particular case our translators are not court certified interpreters so very often the reviewer burden is sometimes high depending on who is doing the initial translation because the translation may be more or less accurate depending on who's doing the volunteer so that's kind of just a quick overview of what the pupils law library volunteer translation project looks like at this time thank you lilan will stop to see if there's any questions on translation memory systems for lilan how do you screen volunteer translators for language ability in the source and target languages so they usually we almost always have volunteers who have educational background in spanish most of our volunteers who are not already attorneys who have who are bilingual in practice in that context there are students who are involved in spanish on in legal classes where they're using spanish in either a clinic setting or in the school settings so that's kind of the way that we do the vetting if they're an attorney and they hold themselves out as being proficient in a foreign language then we rely on their ethical obligation to provide accurate information any other questions everybody clear on how the segmentation works and how this is different than what i think maria will be probably be able to provide a better description of what translation memory is since that's kind of her her daily look there okay well let me pause the screen then any other questions okay so i'm gonna pause and now we're gonna switch over to maria medlin and nicole newman who are both with transcend and they're gonna cover a little bit more on translation memory systems and share from their experience other ways to improve the quality that comes out of these tools so you are now presenters do you want me to run it for you i think we can do it okay thank you this is uh i'm maria medlin and i'm here with nicole and um and we're at transcend and transcend is a language service provider we do certified translation in the u.s minority languages and we also do plain language and nicole is has many hats here at transcend but her biggest hat is she's a software specialist and she handles all of our translation memory well all of our software issues and that includes the translation memory programs that we use and how translation memory integrates with our other software programs and how that impacts languages that use different font systems and have different needs than just the standard weldman alphabet languages and part of my job here at transcend is i work in when i work in other languages i use a translation memory in the editing environment and i also use machine translation mostly for corpora searches to find frequencies of phrasing to determine what's the best choice for different ways of saying things so these are tools that we use here every day at transcend and we'll show you a bit about how we use them on the next slides thank you just really briefly one of our slides i'm sorry that didn't have enough time to get in there is our process and we want to tell you that the way that we do translations is a bit laborious and and we use professional certified translators and then we have them reviewed by professional certified editors and then we have what we call micro proofing so a proofer who goes sentence by sentence to make sure that everything's complete then we have macro proofing so someone who checks in on well they're the same number of elements in the table of contents that's all the address and phone numbers are correct that the key information corresponds and looking more at the global issues and then we have what diana was talking about we have a our clients have a subject matter expert a lawyer from the judicial council who would review our work to make sure that it's legally sufficient and accurate and we make any adjustments that often it leads to a conversation about well let's research this a little more or we'll change this or what do you think about that but we make adjustments as needed we note any adjustments in glossaries so we can make sure that we're tracking our clients preferences you know what the judicial council in california may like maybe a bit different than what Tennessee may like so we would track the preferences by client and then as we've seen field testing is really becoming a reality and that's the ultimate test to make sure that the consumers can read and understand it because if that's not happening then there really wasn't much point in spending all that time and money and doing that translation and so that may lead to further adjustments in the language and in turn that would lead to adjustments in our glossary and make us make some changes on how we're saying things so that's our process so a good question here is what is a certified translator and who does those kind of certifications okay so maybe we could take that i'd be happy to take that first of all there's some information on our website on certification that talks about this all the certifying agencies in the united states that might be one quick way of answering that and if that's okay we could go through the translation memory stuff and come back to that does that sound okay if you can get us a link to that i can get that up to the panel so sure so it's transcend.net library and then look look for a translator certification and so on the next slide this is the type of environment that as an editor that i would work in this is called a two column export file and if you look at the colored areas of the files the translation memory file will tell you how often the the particular tool there's many translation memory tools but how often your memory tool has seen this particular segment before so everywhere it's yellow it's new so this is typical you're doing a new document new topic it will have seen it zero percent times before the ones where it's seen it 100 times before it's like top right you're doing a website it's seen seen top of course many times before so it will naturally fill that in if you're doing if it's the url for the particular website it will have seen that many times before so when it's seen it before it will offer you the same translation that it's seen previously when it's seen it 96 percent times before it could be that maybe something is italicized and that's the change that you have to make or there's one word that's different and that's the change that you have to make so these percentages tag how often you've seen it before and this export file has all the tags in it that also key the text enhancements and so this is one type of file that the tm can generate for you to edit in there's also the more typical type of file which most of the translators prefer and that's on the next slide and this is similar to what lilin was showing us that his translators work in where it goes segment by segment and you can see where it says 100 percent that's seen it before and so it's seen this piece of text this segment 100 percent before and it's going to offer you what the last translator put in and confirmed as the right text before and then if you are the next volunteer translator and you accept that then you just click on it and then go to the next piece and if it hasn't seen that before then it's up to you to translate if you go to the next piece down and it's seen it 100 percent before you can accept it so you just go through piece by piece and then it will reconstruct it and put it together for you in the original format and the last way in which you can edit the text is with your invisibles on so this would be like in a microsoft word document where you click on your paragraph symbol to turn your invisibles on and you just have to be super careful not to do anything to those little tags in there because then you completely mess up the system and then you be sorry you did and it's big pain but if you keep make sure you keep your cursor out of those little tiny red numbers with the less than more than signs then you can see the source text and the target text at the same time the source text if you turned off your invisibles you wouldn't be able to see it you'd only be able to see the Spanish but this is like super helpful for the editors because you can look at the English and Spanish segment by segment at the same time which really speeds you up and then you can as an editor I can make my changes and track changes in what for TM you call an unclean file and then the translator or we can incorporate those changes to memory so that we have what we think is the cleanest most correct version of the translation or let's say that we want like our subject matter reviewer to incorporate her changes too we could also I could put in my changes and then I could send a file like this to the subject matter reviewer after I incorporated my changes and have her do the same so we could incorporate another level of changes so that we could you know we could really get in a lot of level of changes to this file so that we're really spiffing up our memory and incorporating another level of review to it as well so here we're using memory and we're also using the track change tools in Microsoft to be able to improve the translation and to incorporate those corrections into the final TM and to write notes to each other so we use the comment function as well so it could be for instance that in the comment function it we could be writing notes to each other about this is a change I made I'm not really sure but I suggest it for this reason or a note to the client there seems to be a mistake in your English here or you know something like that so it's a really nice way of getting all of your dialogue and your corrections or commentary in the same place and this one I'll go over a little bit but Nicole sitting to my right is the woman who's really our technology expert here too but for the the translation tools one of the things that we found is that some misunderstanding about who the TM is for and who's using it and definitely the TM like who should be using it these are these these are tools for the professional translator and why why should they use it and they should use it because it helps them to be faster if they have documents with repeated text if you're not working with high volume of repeated text and that means like 75 percent and above repetitions it's really not useful to you if you do have 75 percent and above it is useful to you and then of course you have to be able to know how to manage the intricacies of the software and if you're just not that kind of techie person that can do that then it's not for you and then you have to be working with clients that have that sort of need so for instance if you have a client that has every document is really like a new surprise which many clients are like that it doesn't TM is not useful for you but if you have clients that really have a lot of similarities across their documents then it really helps you be faster and it helps you be more efficient if you have clients that have like an average for clients we find that clients in the legal sector tend to have somewhere between like 10 to 12 percent repeated text across their documents so if you have 10 percent repeated text you will see some savings and that's good why not get 10 percent savings it's especially good for form producers you can imagine that there's form producers that have so much repeated text and so those like those little pieces we saw on the German forms it can just pick them up and immediately populate them for you in different areas and also reproduce the formatting which is a huge time savings so that's where it really is of big value and also the term base can be a great tool in terms of keeping glossary standard and this is when it really matters I think to have professional and certified translators to make sure you're using terms that are going to be legally holed up to scrutiny and also that are going to be consistent and standard with what other states and legal service agencies are using and then hopefully as field testing comes more and more into play with what people understand so that when we get to writing articles we're using the same like for expungement we're using the same words that are being used in other states and we're using words that the readers are going to understand so it's not a different word in every state and it's not a word that the readers aren't familiar with and the TM software that most translators use are we just call it Trados and Déjà Vu, MemoQ, Word Fast Pro and Nicole you may know but I don't know people that use the free tools. I don't know anybody that we work with that uses those tools the nice thing about translation memory software it doesn't matter which one they're using the files are interchangeable so once if somebody's using Word Fast Pro they can export the file in a format that you can load into let's say you're using Trados so there is a big benefit to using software that you do have to pay for up front and it's an initial investment but the we because we're a translation company we own Trados but our translators use whatever they want to use and it's on them to have that software update the software so it's not an expenditure on our part any certified professional translator will use their own translation memory cap tools like software so it and then once we get when they send us the file then we load it into our translation memory and we can keep the glossary and term basis updated and all of our segments of translation updated here in the office as well but most most professional certified translators will not use free translation memory software and it's not a huge investment to buy it but it's if you're going to be using it a lot it's worth it definitely to use something that's not free I think you said it for us um I think yeah yeah okay thank you do we have questions on the chat yeah just the one over the certified translators and we dropped a link into the questions for transcends resources there okay well I I do have a question and I want to thank both Maria Nicole and Leland for showing sharing a little bit more background and screens on the translation memory because it's a tool that's been out for a little while but not one that we have really utilized I think that Leland's program is a pioneer in doing that and I I saw that you said that it's really good for high volume you know high volume or people that are using like formulaic formulaic language do you know if courts are using this for their forms would forms be a good candidate for for use with translation memory so it's really it's really a translator tool rather than more of a like a company sort of tool and for you know court forms tend to be I'm speaking in the the forms that we are familiar with tend to be different enough that there's not a high volume of repetition of the text but it's still helpful to use use it because at least you're getting consistency every time you know you see you know defendant plaintive party hearing date that sort of thing so even if you're not getting savings related to efficiencies you are getting consistency now the important thing is what kind of consistency are you getting right so you want to make sure that people that are using your translation memory are certified professionals and are putting the right terms in it because what if you know there you're starting off with well putting that that's I guess what I wanted to say you want the right terms your translation memory is only as good as what's put into it so you have to really make sure that as you build your database and your terms that it's of a certain quality and of a certain level because you're only going to get a translation back out that is I mean and this is basically how google translate and machine translation works as well there's really a lot of similarity in that google translate is only as good as the humans put translation into it and it spits it out using algorithms whereas translation memory tools cat tools is actually produced and controlled by the certified translators so the quality level if it's good quality being put into your translation memory you're going to get a very high quality level out so it's taking it to a next level then like you know we there was a project in Illinois I am probably five years six years ago where they hire somebody that had been trained as an attorney in Mexico City and who was also a certified core translator here in Illinois so he was like triple qualified and he he basically the project was to compare the quality of doing the machine translation the other tool that Diana was talking about and then doing his own translations and the conclusion that came out of that big project what I think the community took out was the the notion that if you're going to use the statistical translation the machine translation that it will always the best practice is then to have a certified translator to review it so these translation memory kind of takes that to a more industrial level right in the sense that and and the tool itself is already built for review and comment like you share so it's a way to doing it more from an enterprise perspective as opposed to just doing it no this is not it doesn't this doesn't assume that you start with machine translation this assumes that you start with a human and that you speed up the human's workflow by recognizing texts that the humans who have worked on this project before have recognizes things that you've translated before so it it there's no machine translation you just feed it the you just feed it the direct content and then a certified interpreter reviews it basically no you give it what you're going to translate you give it the language and you say that what what have we done before and they will give you the matches right that's right it looks at the database and segment by segment and if you've translated it before either on the current document or in the previous database and if you're working on a larger project that other translators have contributed to in that memory then it will use the translations from the memory and offer it to you is it is it easy to share those memories like nationally if we've got a bunch of forms being done in minnesota and we've got a bunch of forms being done in washington can you share those memories does each instance of its own and we're acting in silos well it's a very it's a complicated question and okay and one it has to do with quality control when you share memory it necessarily means that people are adding to the memory and that's something that we handle very cautiously at transcend because we for instance there's we don't use other people's memory if i can answer it that way and so it's possible that you could add things to the memory that would reduce the quality of the memory and then and then you essentially be turning it into mt rather than tm and then there'd be no point in having tm does that make sense i see what you're saying yes so okay i don't if there are other questions let's see if there but i do have a follow-up question for nicole yeah any questions sarke from the audience so have you considered using something like tensor flow to enable machine learning to use translations provided by translators to offer users some automatic translation so instead of taking um general using kind of a curated data set this was from uh david rodriguez is the question for leland um it did not say who the question was for it was just submitted into the question box i think i think that he's asking about uh some data analytics ai tool tensor flow i don't remember maybe we cannot meet david so he can explain what that is because that's um that's a different sort of tool in a in a more advanced ai in quotations i don't remember who makes it i don't remember if it's microsoft or google i've just been mutated in certified legal translation so you know when we when we use tm what it means is that we're relying on this very clean database of translation that's been produced by certified translators checked by certified editors microd macroad gone through subject matter experts and sometimes field tested so we don't want to introduce text that is below those standards so you know we're in a very difficult um you know much uh more specialized area than many other fields so um okay i i'm not sure that's what the vendors are doing you know what i mean like there's the vendors have the revenue flow to create and maintain cutting-edge technology so probably the the programs that they share they some of them may be using that i don't think that at the legal aid level anybody's working necessarily um on machine learning for language that's a huge that's like the holy grail to get to the simultaneous interpreters that we had in star trek and the and the perfect translator without human intervention that's like we need quantum computing so you know i mean that's like a huge huge problem so and i see that you dropped the article um to the presenters to share that with everybody which is a june through 2010 article um what has been the changes in technology in the last um eight years since this article that i'm about to share with everyone um and what has been the objective quality assessment changes since then from from what i understand the algorithms are getting a lot better and faster right and the processing and all of that is a lot faster but it's the the problem is the computing model it's not a question of algorithms is is what my cognitive um you know the people that that that do cognitive computational linguistics at the national labs what they tell me that the problem is that we're still using like silicon you know carbon based and silicon based computing and that we really need is um a supercomputer that can actually use quantum computing and some groups are making breaks those on that and so they say once we get to something that is so much more advanced and fast then we may be able to break the problem but um so the advancements are all in the algorithms and the the big data analytics like the stuff that david was asking about but not but not on the computing model but you know things are advancing again it maybe five years who knows right there's a lot of money to be made because these impacts and the nice commercial world right like you see all the advertisements for drugs and big big time and commerce in latin america right amazon is very interested in in in doing e-commerce in latin america facebook too google so there this is a race this is a this is a problem that has significantly billions of dollars in revenue so it may be solved but it's not gonna be you know it's it's a it's a billion it's a billion dollar problem do you have a question for maria um that just came in um just going back to the um tms and whether they're stored in a cloud oh yes um so it can be it depends on what you're you know how you're set up um some companies had it where they have um their cat tools where there's a basically a the database bank it's it's uh more of a start it's set up as a server um that's like but that's that is very expensive and you have to have mass quantities of uh translation and it wouldn't be for like a small like just somebody using it like as as a translator so that would be more for the company to set up to use this as a server um what we find more often is that it's it's a buy a translate per translator basis so the answer is yes and no to that one so if if a program has someone coming to them offering to do translations um what are the kind of first three steps that that program should take in kind of making sure that they know what they're doing and then helping them get up with get set up with software and tools to do it effectively well it's not so much that the program uh needs the translation tools although it should um retain access to the memory because if they change vendors they'll want to maintain the memory so that they can apply it to future vendors so having them having the memory files um will be something that they they'll want to say that they'll want to keep because let's say you you know work with us for a year and then for whatever reason you don't want to work with us anymore you'll want to keep that memory because you want your language to be consistent with your next vendor so that would be an important point um to negotiate so some of our clients require that we submit um translation memory files uh on let's say a quarterly basis and so they they hold on to the translation memory so that it you know as Maria said sometimes if they go to another one of their vendors and they want to use that translation memory then they have it to send to their other translator it's more of a translators tool than it is a client tool um and you can keep and and manage it um as the client but what we find more often is that they we just export it um as and then we send it to them and it's all agreed upon that you know they may have certain protocol that we have to follow within our translation memories like tags and stuff and information for those segments and it's really kind of up to the client to determine that what kind of information they want in their translation memories and and then for us to provide it to them um but that's all negotiated with with the client and I think as a client too you want to say um you know do do you are you using translation memory is your team using translation memory and um if if you genuinely have texted so varied that it's not suitable to a translation memory discount then you know then it will be hard to give you a translation memory discount but um your but the software will be able to calculate you know what sort of repetitions you're getting so if the the service is using translation memory then you can ask for a translation memory discount so we've got two more questions here or one's a comment which is I believe Trados allows you to use setup translations on a shared directory to share with multiple translators that was from James Ford and then we've got a question that comes up often which is some websites include Google translate buttons for legal aid organizations that use their website to provide provincial clients information about the types of services they offer and what are the pros and cons of the Google translate button specifically on a website so I think Diana's presentation really responded to the perils of using Google translate and one of the things that I like to tell people is I have a son who works at Google and um sorry as you were mentioning I mean Google translate is so much better than it used to be and with the new neural networks and the advanced algorithms it's a lot better I mean we can really see its improvements but Google does not use Google translate to translate their own websites you know we're a vendor to Google as a translator so Google uses human translators to translate their own websites so it's definitely suitable for so many things and it plays such an important role but Google translating legal websites is not an appropriate use of Google translate technology in our opinion and that would be a mismatch. So only cons no pros. Well Diana covered a little bit of that. Right I think that the I think using Google translate by itself definitely terrible the button does give people notice that there could be some mistakes which is better than not notice but it is still rough and needs a lot of work as it improves over time should definitely be evaluated but that's that's where I am on it that the the notice helps a little bit but honestly if you have critical things like where to pay fines or your legal rights you should be getting high quality translations. Well the problem with the button is that it's a crutch right because the button denotes that it's not something that they're doing just because they're on a pinch for two days it is a crutch that could become long term so the better approach would be to budget and put in your website roadmap and budget when you're going to get your content up in your target languages based on the demographics of the community that you serve and if that takes you know hopefully when you when you start the website you you are planning on that and so it won't be a huge delay I don't know in other communities disclaimers and you know legal depending on where people are coming from those disclaimers won't be appropriate and what you may end up they may not even understand the concept of a disclaimer so what may end up happening is that when that community looks at your at your web page and you have the google translate there the translation could be so poor so poor that they would never ever ever come back again because it's disrespectful or completely unintelligible so what percentage of the problems what percentage of the problems with regards to translation is the courts using archaic language and not fully adopting plain language I ask this because washington state is considering a court rule that would push for everything the court creates to be done in plain language who is the question directed to so several people including diana originally talked about the problems with legal specific language and not using plain language so diana potentially but anyone could answer diana do you want to answer I believe she's on she's muted okay so she's on hi I've been trying to jump in here for a while so thank you for no that's quite all right I understand um so oh well just if I could I'm happy to respond to that um I just wanted to chime in on one issue with regards to the disclaimers I think the disclaimers are really helpful to um for those people who are putting that content out to protect them from potential liability um but they're not really helpful to the people who are reading because if you read something and you and it says and I will say also that um I've seen some people spend a whole lot more time getting a professional translation of their disclaimer so that they can then put up machine translation text and I feel like that is just an unwise allocation of resources but it really doesn't help the reader if they um see a lovely translation of a disclaimer saying well this may or may not be reliable information I mean how are they to know what is reliable or unreliable so I just I feel like that's just not a good solution to this issue um but getting back to the question of well is the problem really the source text I would say yes um and is google translate could that work if all source text were fabulously drafted and plain language and everything else and I would still hesitate I don't I I just feel like without knowing what the full context is and without being able to provide that machine translation um with the full context of what this document is I mean if you're putting in the word exit yeah you're going to get the right thing right but if you're putting in something that actually has some legal weight to it I just feel like it's very risky to rely on machine translation um I'm not saying that tomorrow with the you know who knows right I mean when I was in college we had email for the first time so and Lord knows we've come a long way from there so I have no idea what the future holds with machine translation but I would say that right now it is an extremely risky proposition to rely on it for legally accurate translation do you have a question from Taylor who was wondering about using uh google translate for translating legal referral information so program description office location phone number income requirements etc is that for me for any of the panelists I mean I feel like without knowing exactly what it is that I don't know what to say on that um you could certainly try it and see what happens but um I think it could be uh you know if we're talking about one word like telephone you know um but again I kind of um always like to point people back to the fact that we already have lots of things developed that can be easily adapted for that so for example we have a notice of link available language access services and as part of that notice we have translated into several different languages professionally um you know where to get help um who to contact that kind of step and it can all be adapted for the local setting so um again when you're talking about one word or very basic translation google translate might be just fine but why not use something that's already been developed for that purpose and I I'm gonna share that um transcend has some wonderful legal icons and I think that in a lot of ways um like Diana shared a lot of those kind of instructions could be iconized or done more um in a diagram kind of way so I'm putting in the chat to share the transcend legal icons because I think that with some of the LEP communities um that that may that may go a long way and it may minimize the need to have a words or complicated concepts you know if you can use an icon that denotes like a rest or something that's um a little bit more complicated and that that has some legal weight like Diana says like can you pay so check out the icons because I think that those can go a long way if you're creating content or instructions or a workflow and you could use the icons intermittently now that that is a great point I think that legal publications generally could use a lot more icons yeah and they they they get rid of word or the words will you know there'll be a backup for that so I really encourage people to look at at this library that transcend has shared but also there are others in the legal aid community that are doing um good work with workflows and um and simplification of process so all of these things will will lead to better better print and better content any other questions if if someone has a site that has some previously translated items on it what is the best way for them to go about auditing to see if those were done well the process like what what you how you would walk backwards to do a quality review yep Maria um so you could have you could have a certified reviewer review it I think if because we're talking about legal materials you would probably want to have a court-certified reviewer look at it uh each each print of the if you go to our library and look for uh certification mechanisms there's a list of organizations that explain how you contact the different members but to find someone who's court-certified to look at it and say whether they're willing to certify it as correct or to make suggestions or corrections to improve the document or the entries uh any more questions from the audience this has been a very active question and answer session people are very very interested in this I think that wraps it up we will have a blog post that pulls out some of the major points and we will have the full video up within a week on our youtube channel and we will send that out along with ways to uh find or contact our speakers thank you so much for putting this on pro bono net we will be having a second webinar on language access that covers some different topics a little bit more on the traditional side of language access later on this year anything you would like to close with jillian or claudia I I want to thank our panelists for for for joining and and working on on pulling this content together I hope it's helpful to the audience and that it's giving you some exposure to to translation memory which is not something we have talked a lot in the legal core nonprofit world and hopefully if if anybody's you know give us the feedback in the survey and I am hoping that we will see more groups use the translation memory and to produce content and maybe hopefully in the long term we can dream of collaboration between groups that are using pretty much routine language all around so if anybody wants to think or brainstorm about that feel free to to reach out but this this has been very educational for me to see the back end of the translation memory and I hope everybody has enjoyed the exposure to those um those capacities that that are we are fairly new and we haven't used and just a plug that some of that could also be used to translate to plain language just some some plant a seed on that too excellent thank you so much claudia thank you to all the panelists we greatly appreciate it we we look forward to doing much more on this topic as the years go on this is one of our most popular webinars thank you thank you thank you